Subject: Re: Rhythm perception From: Henkjan Honing <honing(at)UVA.NL> Date: Tue, 15 Nov 2005 15:35:51 +0100On 15 Nov 2005, at 15:27, Dan Ellis wrote: >>> We were discussing rhythm patterns the other day and the >>> question came >>> up about how one determined the start of a pattern. > > I think this is the "downbeat detection" problem, which is distinct > from > the problem of deciding the main repetitive period of the rhythm > i.e. the spacing between downbeats. "beat induction" (footapping, beat tracking, tempo tracking etc) has (at least) two components: phase and period. Longuet-Higgins & Lee (1984) determine both of a given rhythm, Large & Kolen do a similar thing using a couple oscillator model. Ed with probably join is as well ...:-) Henkjan > I'm not a musicologist, but I've > tried to build a system to align downbeats in different rhythm > patterns, > so I've thought about it a little. > > Here's one piece of evidence: Tristan Jehan presented a poster at the > recent Mohonk WASPAA on a machine learning approach for predicting > the downbeats, which he tested on a style of Brazilian music called > maracatu. He played the maracatu examples to listeners familiar with > northern-hemisphere western music, and they could not identify the > downbeat - but of course maracatu fans have no problem indicating the > appropriate downbeat. > > http://web.media.mit.edu/~tristan/Papers/WASPAA05_Tristan.pdf > > That to me is powerful evidence that downbeat is a cultural norm > and not > strongly based on some universal. > > DAn. ************************************ Henkjan Honing E honing(at)uva.nl I http://www.hum.uva.nl/mmm/hh/ Music Cognition Group I http://www.hum.uva.nl/mmm/ F feed://www.hum.uva.nl/mmm/news.rss ************************************